Gameplay Video Event Detection Using Perceptual Hashing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

It is difficult for users to simultaneously play video games and record exciting or surprising content while also summarizing game progress, as identifying significant events within long gameplay sessions is challenging.

Innovation Solution

An apparatus and method for video tracking that uses perceptual hashing to detect scene cuts and identify notable events in video game footage, employing image reduction, hash generation, and database comparison to efficiently identify and tag significant game moments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If users manually monitor and record gameplay footage, then they can capture exciting or surprising content, but they cannot simultaneously play the game and selectively record footage

Engineering Contradiction:
Improveautomated event detectionVSAvoidsystem complexity
Core Design Contradiction:
Extent of automationVSDevice complexity

Solution Approach 1:

The patent replaces manual mechanical monitoring with automated perceptual hashing algorithms. The system automatically generates perceptual hashes of game screenshots, compares them against a database of event hashes, and identifies significant moments without requiring manual user intervention, thereby achieving automated event detection while maintaining manageable system complexity

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent creates perceptual hash copies of game footage frames. Instead of storing and analyzing entire high-resolution video frames, the system generates compact perceptual hash representations that can be quickly compared against event databases, enabling efficient automated detection without processing the full complexity of original game footage

Inventive Principle:
Principle #26Copying

2Loss of information

If users record all gameplay footage, then they have complete game progress, but it becomes difficult to identify significant events within many hours of content

Engineering Contradiction:
Improveevent identification accuracyVSAvoidtime to review footage
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The patent extracts only the perceptual hash features from complete gameplay footage rather than analyzing entire video files. By extracting these distinctive visual fingerprints and comparing them against event databases, the system identifies significant moments without requiring users to review hours of footage, thus preserving event identification accuracy while dramatically reducing time loss

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent performs preliminary action by pre-computing and storing perceptual hashes of significant game events in a database before gameplay occurs. During gameplay, the system only needs to generate current frame hashes and compare them against this pre-prepared database, enabling rapid event identification without time-consuming analysis of recorded footage

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If the system analyzes every frame of gameplay video, then it can identify all events, but computational overhead becomes excessive

Engineering Contradiction:
Improveevent detection accuracyVSAvoidcomputational resources
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent uses perceptual hash copies as simplified representations of game frames. Instead of performing complex image analysis on every full-resolution frame, the system generates compact hash copies that capture essential visual information, enabling accurate event detection while minimizing computational resource consumption

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent applies partial action by analyzing only the perceptual hash representation of each frame rather than the complete image data. This partial analysis approach maintains sufficient event detection accuracy while dramatically reducing the computational overhead associated with processing full-resolution video frames

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4704050A1Apparatus and method of video tracking
Publication Date: 2026.03.04 SONY INTERACTIVE ENTERTAINMENT LLC
  • EP4704050A1 patent drawingFigure 1
  • EP4704050A1 patent drawingFigure 2A~2B
  • EP4704050A1 patent drawingFigure 2C~2D

AI summary

A method of identifying an event that is graphically acknowledged within a sequence of images, comprises the steps of defining one or more regions of respective images in which a graphical indication of acknowledgement will occur, selecting at least a first defined region as a candidate acknowledgement image, and selecting for output one or more of the candidate acknowledgement images, or respective images of the sequence of images that comprise such candidate acknowledgement images.